CN104469879A - Dynamic k value clustering routing method - Google Patents

Abstract

The invention provides a dynamic k value clustering routing method. According to the method, the clustering structure of a network is partitioned in the first round of each period, the effective k value of the network is calculated according to the current network clustering situation and energy distribution, the clustering structure is kept unchanged in the following k rounds, and only cluster head updating is carried out in a cluster. The k value will be recalculated at the beginning time of each period, and therefore the number of the rounds of the network periods is dynamically changed; influences of poor clustering on the performance of the network can be effectively lowered through dynamical setting of the network clustering periods, the energy utilization rate of network nodes can be improved, and loads of nodes in the cluster can be effectively balanced.

Description

A kind of dynamically k value cluster routing method

Technical field

The present invention relates to Cluster-Based Routing Protocols for Wireless Sensor field, particularly a kind of dynamically k value cluster routing method.

Background technology

In wireless sensor network, distribution sensor node is in a network responsible for gathering the Various types of data in environment, and data are periodically sent to base station.Due to sensor node finite energy, each operation gathering and send data all can consume certain energy.Along with the operation of network, increasing node is dead due to depleted of energy, causes the performance of network constantly to decline and even paralyses.In traditional wireless sensor network, node is all transmit the data of collection by directly carrying out communicating with base station, and energy consumption size and node to base station distance be directly proportional, this process need lot of energy.In order to make node energy more for data acquisition, need to introduce Routing Protocol in wireless sensor network, this also becomes a study hotspot in wireless senser field.

At present existingly a lot carried out based on the energy-saving routing protocol of wireless sensor network, substantially they can be divided into plane Routing Protocol and clustering route protocol from the angle of network topology structure.In plane Routing Protocol, the status of each network node is equality, and they generate route by partial operation each other and information feed back.In this quasi-protocol, destination node sends querying command to the node of monitored area, after the node in monitored area receives querying command, sends Monitoring Data to destination node.Plane Routing Protocol due to each node equal, so network configuration simply, is easily expanded, not easily produce bottleneck effect, but owing to lacking management node, thus the optimization to communication data is lacked, self-organizing collaborative work algorithm is also very complicated, and comparatively slow to network dynamic change reaction speed, typical plane Routing Protocol has SPIN, SAR etc.

Plane of comparing Routing Protocol, the network in clustering route protocol is divided into multiple bunches, and each bunch is made up of ordinary node and leader cluster node.Ordinary node completes the collection of data and data is mail to a bunch head, leader cluster node to be responsible for bunch in the management of member node and data message and data bunch between forward, in addition bunch head can also to coordinate bunch in work between member node.The advantage of clustering route protocol is that bunch head has merged data that node sends and then forwarded, and can greatly reduce the traffic in network, member node function is fairly simple, need not maintaining routing information.Simultaneously with plane road by compared with, more easily overcome sensor node and move the problem brought.

Clustering route protocol can be divided into single-hop clustering route protocol and multi-hop clustering route protocol according to the difference of data transfer mode.The typical feature of single-hop clustering route protocol is: bunch head in network, after collecting the data that each member node sends, directly can send to base station by after data fusion.And in the network using multi-hop clustering route protocol, leader cluster node sends to last layer leader cluster node after the data that member node is sent being merged, data can via the path transmission designed to base station.The typical feature of multi-hop clustering route protocol is, leader cluster node is not only responsible for the transfer of data in this bunch, is also responsible for the transfer of data of lower level node bunch simultaneously.

In the network that have employed single-hop clustering route protocol, the node division of labor in network is clear and definite: ordinary node is responsible for collecting data and is sent the data to the leader cluster node at place bunch, leader cluster node be responsible for receive bunch in member node gather data, and directly send to base station by after data fusion, simultaneously when local choosing bunch, be responsible for selecting new bunch of head, then bunch in broadcast packet containing the message packet of gap information and new bunch of header.

LEACH-C agreement is typical single-hop clustering route protocol.In the network using LEACH-C agreement, base station collect at every turn one take turns the data that all nodes collect after, can sub-clustering again be carried out.Again, during sub-clustering, the information of self must be sent to base station by each node in network, the direct shortcoming of this agreement be exactly frequently clustering operation cause node to consume unnecessary energy.Have researcher to propose LEACH-EB agreement, in LEACH-EB agreement, network cycle expands to 1+k wheel for this reason.Within each cycle, the first round carries out a bunch head and selects and divide network cluster dividing structure, then keeps cluster structured constant at ensuing k wheel, and in each bunch, only carry out bunch first watch new.Effectively can reduce the number of times of network cluster dividing operation in this way, thus extend Network morals.But the k value in LEACH-EB agreement is changeless, owing in infonnation collection process interior joint Energy distribution being variation in real time, the cluster structured of each cycle is change simultaneously, adopts fixed k to be difficult to the demand meeting network dynamic change.

Summary of the invention

The present invention is directed to the deficiency of existing LEACH-EB agreement, propose a kind of dynamically k value cluster routing method.

Technical scheme of the present invention is: a kind of dynamically k value cluster routing method, comprises the steps:

Step 1, at deployment base station, monitored area and sensor node, sensor node is started working and self-organizing forms wireless sensor network;

Step 2, all nodes in network send the packet comprising self-position and dump energy information to base station; After the packet of all nodes is collected in base station, calculate the dump energy mean value E of surviving node in current network _average, and select residue energy of node higher than E _averagenode alternatively bunch head, generate candidate cluster head set;

Step 3, base station solves best sub-clustering mode, calculates the effective k value of rational network simultaneously, and by these information broadcastings;

Step 4, after the node in network receives the sub-clustering packets of information of base station broadcast, searches bunch head ID at self place bunch; If bunch head ID at self ID and place bunch is identical, then node self is elected as a bunch head, and node travels through a sub-clustering message bag according to bunch head ID simultaneously, the id information of all member node in collection bunch; If node finds certainly as ordinary node after receiving sub-clustering packets of information, so node enters wait state, until receive bunch head broadcast message at place bunch, the distribution time slot of each node in wherein comprising bunch, bunch interior nodes obtains the time slot that self sends data from message packet;

Step 5, enter the stabilization sub stage of transfer of data, bunch interior nodes sends bunch head of data message to place bunch of themselves capture at assigned timeslot, after bunch head receives the data message that all member node send, data are carried out merging and forwarding give base station, after the stabilization sub stage terminates, go to step 6;

Step 6, network has completed the 1st of one-period and has taken turns, and enters local subsequently and selects manifold flow journey, namely keeps cluster structured constant at ensuing k wheel, and in each bunch, only carry out bunch first watch new.7 are gone to step after local choosing bunch terminates;

Step 7, judges that whether nodes is all dead, is terminate current process, otherwise goes to step 2.

As preferably, in described step 3, base station traversal candidate cluster head set, based on member node in all bunches to bunch head square distance and minimum principle, adopts simulated annealing to solve the sub-clustering mode of the best.

As preferably, in described step 3, calculate an effective k value of rational network according to network cluster dividing situation and node energy distribution, the computational methods of described rational network effective k value are as follows:

Suppose there be m bunch in network after the 1st polling bunch completes, bunch numbering be respectively C ₁, C ₂..., C _m, the node number in each bunch is respectively N ₁, N ₂..., N _m, leader cluster node numbering is respectively CH ₁, CH ₂..., CH _m, for the i-th bunch, N in bunch _ithe numbering of individual node is respectively C _i1, C _i2..., in order to the effective k value of computing network, this method introduces the concept of " parallel bunch of head ", " in bunch effective k value " and " the effective k value of network ".

This method serves as the ability of bunch head according to formula (1) computing node;

$T\left(C_{\mathrm{ij}}\right)=\frac{E\left(C_{\mathrm{ij}}\right)}{p\×D_{C_{\mathrm{ij}}\→\mathrm{BS}}^2+(1-p)\×D_{C_{\mathrm{ij}}\→\mathrm{others}}}---\left(1\right)$

Wherein C _ijthe node of j is numbered, T (C in representing the i-th bunch _ij) represent node C _ijability weights, E (C _ij) represent node C _ijdump energy, represent node C _ijto the distance of base station, represent node C _ijto bunch in the quadratic sum of other nodal distances, balance factor p be used for adjustment node to base station distance and node bunch in position node served as to the impact of bunch head, the value of p, between 0 to 1, has simultaneously:

$D_{C_{\mathrm{ij}}\→\mathrm{others}}=\underset{h=1}{\overset{N_i}{\mathrm{\Σ}}}{D}_{C_{\mathrm{ij}}\→C_{\mathrm{ih}}}^2---\left(2\right)$

Wherein the node being numbered j in representing the i-th bunch is to the distance of node being numbered h.

Define 1. parallel bunches of heads: if node C _ijability weights be more than or equal to bunch head CH at place bunch _iability weights T (CH _i), then this node parallel bunch of head being bunch, namely has T (C _ij)>=T (CH _i).

Effective k value in defining 2. bunches: in network single bunch can carry out local bunch first watch new number of times be called this bunch bunch in effective k value, if bunch C _iparallel bunch of head number be z, so have k (C _i)=z.

The parallel bunch of head of in network each bunch is found out, existing hypothesis bunch C according to definition 1 _iparallel bunch of head number be z, so bunch C _ican carry out local bunch first watch new number of times is z, this method one bunch can be carried out local bunch first watch new number of times as this bunch bunch in effective k value.

When the node of ability maximum weight during bunch head is bunch, the value of z is 0; This method is provided with a lower limit x during effective k value in compute cluster, improvement bunch in effectively k value calculating method is as follows:

k(C _i)＝max{x,z}(2≤x≤5) (3)

Wherein lower limit x gets the random integers between 2 to 5.

Define 3. network effective k value: in network in one-period all bunches carry out new number of times on bunch first watch, be designated as NK.

By formula (3) to obtain in network each bunch bunch in effective k value, this method using bunch in all nodes to the mean value of base station distance as bunch to the distance of base station, by the ratio square to base station distance maximum bunch in the distance and network of base station bunch, as bunch in effectively k value relative to the weight factor of the effective k value of network.Bunch to the computational methods of base station distance as shown in formula (4):

$D\left(C_i\right)=\frac{\underset{j=1}{\overset{N_i}{\mathrm{\Σ}}}{D}_{C_{\mathrm{ij}}\→C_{\mathrm{BS}}}}{N_i}\left(4\right)$

Wherein N _irepresent the node number of bunch i, represent node C _ijto the distance of base station.

Bunch C _ithe computational methods of weight factor are as shown in formula (5);

$\mathrm{WF}\left(C_i\right)=\frac{D^2\left(C_i\right)}{{D^2}_{\max }}---\left(5\right)$

Wherein D (C _i) represent bunch C _ito the distance of base station, D _maxto represent in network bunch the maximum to base station distance.

In conjunction with formula (3), (4) and (5), the computing formula deriving network effective k value is as follows:

$\mathrm{NK}=\frac{\underset{i=1}{\overset{m}{\mathrm{\Σ}}}\mathrm{WF}\left(C_i\right)\×k\left(C_i\right)}{m}\left(6\right)$

Wherein WF (C _i) represent bunch C _iweight factor, k (C _i) represent bunch C _iparallel bunch of head number.

As preferably, described p span is 0.7 to 0.9.

As preferably, described p value is 0.8.

As preferably, in step 6, described local bunch head renewal process, comprises the steps:

Step 6.1, members list in former bunch of head traversal bunch, according to the ability weights of formula (1) and (2) calculating member node, the node of selective power maximum weight, as new bunch of head, goes to step 6.2;

Step 6.2, former bunch of head broadcasts new bunch of header and gap information, and in bunch, member upgrades corresponding information, then enters the transfer of data stabilization sub stage, goes to step 6.3;

Step 6.3, judges whether the local bunch head renewal rewards theory completing k wheel, is go to step 7, otherwise goes to step 6.1.

Technique effect of the present invention is: a kind of dynamically k value cluster routing method, take turns in the 1st of each cycle and divide the cluster structured of network, then an effective k value of network is calculated in conjunction with current network sub-clustering situation and Energy distribution, and in ensuing k wheel, keep cluster structured constant, bunch in carry out the renewal of bunch head.K value can recalculate when each cycle starts, and therefore the wheel number of network cycle is dynamic change; By dynamically arranging the network cluster dividing cycle, effectively can reduce the bad cluster structured impact on network performance, improving the capacity usage ratio of network node, the load of efficient balance bunch interior nodes.

Accompanying drawing explanation

Fig. 1 is dynamic k value cluster routing method flow chart.

Embodiment

Existing LEACH-EB agreement decreases by being become by the network operation period expansion 1+k to take turns the number of times that information is collected in base station, extend Network morals to a certain extent, but changeless k value is difficult to the demand meeting network dynamic change.The present invention provides a kind of dynamically k value cluster routing method accordingly.Taking turns in the 1st of each cycle adopts LEACH-C agreement to divide network cluster dividing structure, and draw an effective k value of network according to the dynamic k value calculating method that the present invention proposes, then in ensuing k wheel, keep cluster structured constant, bunch in carry out the renewal of bunch head.Wherein, LEACH-C agreement is carried out bunch head based on node location and dump energy and is selected, and this mode makes network cluster dividing structure more reasonable.Meanwhile, the dynamic k value calculating method that the present invention proposes is cluster structured and Energy distribution Network Based, and this makes the network operation cycle can adjust dynamically along with network operation situation.Technical scheme of the present invention is described in detail below in conjunction with embodiment.

The wireless sensor network of the present invention's research is distributed in the square area of a M × M, and formed by the sensor node self-organizing of N number of random distribution, its application scenarios is periodic data acquisition, specifically describes as follows:

(1) base station is positioned at outside observation area, all nodes and base station after deployment all no longer occurrence positions move;

(2) all node isomorphisms and primary power is equal, possess data fusion function;

(3) each node has unique identify label (ID);

(4) all nodes can both one jump to and reach base station;

(5) according to the distance of recipient, node freely can adjust transmitted power to save energy consumption;

(6) link is symmetrical.If known the other side's transmitted power, then node can calculate sender and the distance of self according to received signal strength.

In the network of the present invention's research, there is the element of two types: node and base station.All nodes all have two kinds of patterns: ordinary node pattern and leader cluster node pattern.When node is in ordinary node pattern, is responsible for perception information and the information perceived is sent to bunch head at place bunch.When node is in leader cluster node pattern, the perception information that sends of ordinary node in being responsible for receiving bunch, and information is carried out merging then sending to base station, node also bears the task of selecting new bunch of head when local choosing bunch simultaneously.In the entire network, the selection of the base stations being in charge first round bunch head and the division of network cluster dividing structure, be also responsible for the perception information that reception bunch hair send simultaneously.

As shown in Figure 1, concrete steps are as follows for dynamic k value cluster routing method flow chart:

Step 1, at deployment base station, monitored area and sensor node, sensor node is started working and self-organizing forms wireless sensor network, goes to step 2;

Step 2, all nodes in network send the packet comprising self-position and dump energy information to base station; After the packet of all nodes is collected in base station, calculate the dump energy mean value E of surviving node in current network _average, and select residue energy of node higher than E _averagenode alternatively bunch head, generate candidate cluster head set, go to step 3;

Step 3, base station solves best sub-clustering mode, calculates the effective k value of rational network simultaneously, and by these information broadcastings.Go to step 4;

First, base station traversal candidate cluster head set, based on member node in all bunches to bunch head square distance and minimum principle, simulated annealing is adopted to solve best sub-clustering mode, and calculate an effective k value of rational network according to network cluster dividing situation and node energy distribution, then sub-clustering information and k value are carried out the whole network broadcast, wherein the effective k value of network calculates according to formula (3), (4), (5) and (6).

(1) calculating of the effective k value of network

Suppose there be m bunch in network after the 1st polling bunch completes, bunch numbering be respectively C ₁, C ₂..., C _m, the node number in each bunch is respectively N ₁, N ₂..., N _m, leader cluster node numbering is respectively CH ₁, CH ₂..., CH _m, for the i-th bunch, N in bunch _ithe numbering of individual node is respectively C _i1, C _i2..., in order to the effective k value of computing network, this method introduces the concept of " parallel bunch of head ", " in bunch effective k value " and " the effective k value of network ".

Research finds, node whether can serve as bunch head mainly by self rest energy, node bunch in position and impact in node to base station distance three, this method devises formula (1) and carrys out the ability that computing node serves as bunch head for this reason.

$T\left(C_{\mathrm{ij}}\right)=\frac{E\left(C_{\mathrm{ij}}\right)}{p\×D_{C_{\mathrm{ij}}\→\mathrm{BS}}^2+(1-p)\×D_{C_{\mathrm{ij}}\→\mathrm{others}}}---\left(1\right)$

Wherein C _ijthe node of j is numbered, joint T (C in representing the i-th bunch _ij) represent some C _ijability weights, E (C _ij) represent the dump energy of node, represent the distance of node to base station, represent node to bunch in the quadratic sum of other nodal distances, p is balance factor, has simultaneously:

$D_{C_{\mathrm{ij}}\→\mathrm{others}}=\underset{h=1}{\overset{N_i}{\mathrm{\Σ}}}{D}_{C_{\mathrm{ij}}\→C_{\mathrm{ih}}}^2---\left(2\right)$

Wherein the node being numbered j in representing the i-th bunch is to the distance of node being numbered h.

Balance factor p be mainly used for adjustment node to base station distance and node bunch in position node served as to the impact of bunch head, the value of p is between 0 to 1.The energy consumption that node sends data to base station observable index when serving as bunch head collects bunch interior nodes image data is much bigger.Found by Multi simulation running experiment, time between p value 0.7 to 0.9, network operation state is more excellent, and it is 0.8 that the present invention gets p value.

Define 1. parallel bunches of heads: if node C _ijability weights be more than or equal to bunch head CH at place bunch _iability weights T (CH _i), then this node parallel bunch of head being bunch, namely has T (C _ij)>=T (CH _i).

Effective k value in defining 2. bunches: in network single bunch can carry out local bunch first watch new number of times be called this bunch bunch in effective k value, if bunch C _iparallel bunch of head number be z, so have k (C _i)=z.

The parallel bunch of head of in network each bunch can be found out, existing hypothesis bunch C according to definition 1 _iparallel bunch of head number be z, so bunch C _ican carry out local bunch first watch new number of times is z, this method one bunch can be carried out local bunch first watch new number of times as this bunch bunch in effective k value.

When the node of ability maximum weight during bunch head is bunch, the value of z is 0; The ratio accounting for node total number due in network bunch of head number is generally about 5%, when one bunch bunch in effectively k value is 0 time, cause larger impact by effective k value calculating of whole network; Therefore this method is provided with a lower limit x during effective k value in compute cluster, improvement bunch in effectively k value calculating method is as follows:

k(C _i)＝max{x,z}(2≤x≤5) (3)

Wherein lower limit x gets the random integers between 2 to 5.

Define 3. network effective k value: in network in one-period all bunches carry out new number of times on bunch first watch, be designated as NK.

By formula (3) can to obtain in network each bunch bunch in effective k value, in order to calculate an effective k value of rational network, being evenly distributed of hypothetical network interior joint, the distance bunch arriving base station is far away, and the load of correspondence bunch head is also larger.Network effective k value is larger, and network cycle also increases thereupon, from base station more away from bunch also more at the energy of network single cycle internal consumption, therefore from base station more away from bunch also more responsive to network effective k value.This method using bunch in all nodes to the mean value of base station distance as bunch to the distance of base station, and to the ratio square of base station distance maximum bunch in the distance and network of base station bunch, as bunch in effective k value relative to the weight factor of the effective k value of network.What formula (4) described is bunch computational methods arriving base station distance.

$D\left(C_i\right)=\frac{\underset{j=1}{\overset{N_i}{\mathrm{\Σ}}}{D}_{C_{\mathrm{ij}}\→C_{\mathrm{BS}}}}{N_i}\left(4\right)$

Wherein N _irepresent the node number of bunch i, represent node C _ijto the distance of base station.

Bunch C _ithe computational methods of weight factor are as shown in formula (5);

$\mathrm{WF}\left(C_i\right)=\frac{D^2\left(C_i\right)}{{D^2}_{\max }}---\left(5\right)$

Wherein D (C _i) represent bunch C _ito the distance of base station, D _maxto represent in network bunch the maximum to base station distance.

In conjunction with formula (3), (4) and (5), the computing formula can deriving network effective k value is as follows:

$\mathrm{NK}=\frac{\underset{i=1}{\overset{m}{\mathrm{\Σ}}}\mathrm{WF}\left(C_i\right)\×k\left(C_i\right)}{m}\left(6\right)$

Wherein WF (C _i) represent bunch C _iweight factor, k (C _i) represent bunch C _iparallel bunch of head number.

Step 4, after the node in network receives the sub-clustering packets of information of base station broadcast, searches bunch head ID at self place bunch; If bunch head ID at self ID and place bunch is identical, then node self is elected as a bunch head, and node travels through a sub-clustering message bag according to bunch head ID simultaneously, the id information of all member node in collection bunch; If node finds certainly as ordinary node after receiving sub-clustering packets of information, so node enters wait state, until receive bunch head broadcast message at place bunch, and the distribution time slot of each node in wherein comprising bunch, bunch interior nodes obtains the time slot that self sends data from message packet, goes to step 5;

Step 5, enter the stabilization sub stage of transfer of data, bunch interior nodes sends bunch head of data message to place bunch of themselves capture at assigned timeslot, after bunch head receives the data message that all member node send, data are carried out merging and forwarding give base station, after the stabilization sub stage terminates, go to step 6;

Step 6, network has completed the 1st of one-period and has taken turns, and enters local subsequently and selects manifold flow journey, namely will keep cluster structured constant at ensuing k wheel, and in each bunch, to carry out bunch first watch new.7 are gone to step after local choosing bunch terminates;

Local New Policy on bunch first watch:

When local bunch first watch is new, members list in former bunch of head traversal bunch, according to the ability weights of surviving node in the dump energy of node and position calculation bunch, the node of selective power maximum weight is as new bunch of head.If former bunch of head is elected as new bunch of head, then self member node list new on former bunch of first watch, rejects the node of dead, surviving node distribution time slot in then again be bunch, and bunch in broadcast this gap information; Otherwise if former bunch of head is not new bunch of head, then former bunch of head adopts the time slot of new bunch of head as self time slot, then broadcast packet containing new bunch of head ID and bunch in the message packet of surviving node ID.When bunch in member node receive information after judge whether self is new bunch of head, if it is upgrade self bunch header, traversal message packet to set up bunch in member node list, then enter data transfer phase; If node is not a bunch head, then upgrades self bunch header and according to the time slot of time slot Receive message self, then enter the transfer of data stabilization sub stage.

Step 6 is local bunch head renewal process, can be divided into following little step:

Step 6.1, members list in former bunch of head traversal bunch, according to the ability weights of formula (1) and (2) calculating member node, the node of selective power maximum weight, as new bunch of head, goes to step 6.2;

Step 6.2, former bunch of head broadcasts new bunch of header and gap information, and in bunch, member upgrades corresponding information, then enters the transfer of data stabilization sub stage, goes to step 6.3;

Step 6.3, judges whether the local bunch head renewal rewards theory completing k wheel, is go to step 7, otherwise goes to step 6.1;

Step 7, judges that whether nodes is all dead, is terminate current process, otherwise goes to step 2.

Specific embodiment described in the invention is only to the explanation for example of the present invention's spirit.Those skilled in the art can do various amendment or supplement or adopt similar mode to substitute to described specific embodiment, but can't depart from spirit of the present invention or surmount the scope that appended claims defines.

Claims (6)

1. a dynamic k value cluster routing method, is characterized in that, comprise the steps:

Step 1, at deployment base station, monitored area and sensor node, sensor node is started working and self-organizing forms wireless sensor network;

Step 2, all nodes in network send the packet comprising self-position and dump energy information to base station; After the packet of all nodes is collected in base station, calculate the dump energy mean value E of surviving node in current network _average, and select residue energy of node higher than E _averagenode alternatively bunch head, generate candidate cluster head set;

Step 3, base station solves best sub-clustering mode, calculates the effective k value of rational network simultaneously, and by these information broadcastings;

Step 4, after the node in network receives the sub-clustering packets of information of base station broadcast, searches bunch head ID at self place bunch; If bunch head ID at self ID and place bunch is identical, then node self is elected as a bunch head, and node travels through a sub-clustering message bag according to bunch head ID simultaneously, the id information of all member node in collection bunch; If node finds certainly as ordinary node after receiving sub-clustering packets of information, so node enters wait state, until receive bunch head broadcast message at place bunch, the distribution time slot of each node in wherein comprising bunch, bunch interior nodes obtains the time slot that self sends data from message packet;

Step 5, enter the stabilization sub stage of transfer of data, bunch interior nodes sends bunch head of data message to place bunch of themselves capture at assigned timeslot, after bunch head receives the data message that all member node send, data are carried out merging and being transmitted to base station, after the stabilization sub stage terminates, goes to step 6;

Step 6, network has completed the 1st of one-period and has taken turns, and enters local subsequently and selects manifold flow journey, namely will keep cluster structured constant at ensuing k wheel, and in each bunch, to carry out bunch first watch new; 7 are gone to step after local choosing bunch terminates;

Step 7, judges that whether nodes is all dead, is terminate current process, otherwise goes to step 2.

2. one according to claim 1 dynamic k value cluster routing method, it is characterized in that: in described step 3, base station traversal candidate cluster head set, based on member node in all bunches to bunch head square distance and minimum principle, adopts simulated annealing to solve the sub-clustering mode of the best.

3. one according to claim 1 dynamic k value cluster routing method, it is characterized in that, in described step 3, calculate an effective k value of rational network according to network cluster dividing situation and node energy distribution, the computational methods of described rational network effective k value are as follows:

Suppose there be m bunch in network after the 1st polling bunch completes, bunch numbering be respectively C ₁, C ₂..., C _m, the node number in each bunch is respectively N ₁, N ₂..., N _m, leader cluster node numbering is respectively CH ₁, CH ₂..., CH _m, for the i-th bunch, N in bunch _ithe numbering of individual node is respectively C _i1, C _i2..., in order to the effective k value of computing network, this method introduces the concept of " parallel bunch of head ", " in bunch effective k value " and " the effective k value of network ";

This method serves as the ability of bunch head according to formula (1) computing node;

$T\left(C_{\mathrm{ij}}\right)=\frac{E\left(C_{\mathrm{ij}}\right)}{p\×D_{C_{\mathrm{ij}\→\mathrm{BS}}}^2+(1-p)\×D_{C_{\mathrm{ij}}\→\mathrm{others}}}---\left(1\right)$

Wherein C _ijthe node of j is numbered, joint T (C in representing the i-th bunch _ij) represent some C _ijability weights, E (C _ij) represent the dump energy of node, represent the distance of node to base station, represent node to bunch in the quadratic sum of other nodal distances, p is balance factor, has simultaneously:

$D_{C_{\mathrm{ij}}\→\mathrm{others}}=\underset{h=1}{\overset{N_i}{\mathrm{\Σ}}}{D}_{C_{\mathrm{ij}}\→C_{\mathrm{ih}}}^2---\left(2\right)$

Wherein the node being numbered j in representing the i-th bunch is to the distance of node being numbered h; Balance factor p be used for adjustment node to base station distance and node bunch in position node served as to the impact of bunch head, the value of p is between 0 to 1;

Define 1. parallel bunches of heads: if node C _ijability weights be more than or equal to bunch head CH at place bunch _iability weights T (CH _i), then this node parallel bunch of head being bunch, namely has T (C _ij)>=T (CH _i);

Effective k value in defining 2. bunches: in network single bunch can carry out local bunch first watch new number of times be called this bunch bunch in effective k value, if bunch C _iparallel bunch of head number be z, so have k (C _i)=z;

The parallel bunch of head of in network each bunch is found out, existing hypothesis bunch C according to definition 1 _iparallel bunch of head number be z, so bunch C _ican carry out local bunch first watch new number of times is z, this method one bunch can be carried out local bunch first watch new number of times as this bunch bunch in effective k value;

When the node of ability maximum weight during bunch head is bunch, the value of z is 0; This method is provided with a lower limit x during effective k value in compute cluster, improvement bunch in effectively k value calculating method is as follows:

k(C _i)＝max{x,z}(2≤x≤5) (3)

Wherein lower limit x gets the random integers between 2 to 5;

Define 3. network effective k value: in network in one-period all bunches carry out new number of times on bunch first watch, be designated as NK;

By formula (3) to obtain in network each bunch bunch in effective k value, this method using bunch in all nodes to the mean value of base station distance as bunch to the distance of base station, by the ratio square to base station distance maximum bunch in the distance and network of base station bunch, as bunch in effectively k value relative to the weight factor of the effective k value of network; Bunch to the computational methods of base station distance as shown in formula (4);

$D\left(C_i\right)=\frac{\underset{j=1}{\overset{N_i}{\mathrm{\Σ}}}{D}_{C_{\mathrm{ij}}\→C_{\mathrm{BC}}}}{N_i}---\left(4\right)$

Wherein represent node C _ijto the distance of base station;

Bunch C _ithe computational methods of weight factor are as shown in formula (5);

$\mathrm{WF}\left(C_i\right)=\frac{D^2\left(C_i\right)}{{D^2}_{\max }}---\left(5\right)$

Wherein D (C _i) represent bunch C _ito the distance of base station, D _maxto represent in network bunch the maximum to base station distance;

In conjunction with formula (3), (4) and (5), the computing formula deriving network effective k value is as follows:

$\mathrm{NK}=\frac{\underset{i=1}{\overset{m}{\mathrm{\Σ}}}\mathrm{WF}\left(C_i\right)\×k\left(C_i\right)}{m}---\left(6\right)$

Wherein WF (C _i) represent bunch C _iweight factor, k (C _i) represent bunch C _iparallel bunch of head number.

4. one according to claim 3 dynamic k value cluster routing method, is characterized in that: described p span is 0.7 to 0.9.

5. one according to claim 3 dynamic k value cluster routing method, is characterized in that: described p value is 0.8.

6. one according to claim 1 dynamic k value cluster routing method, is characterized in that, in step 6, described local bunch head renewal process, comprises the steps:

Step 6.1, members list in former bunch of head traversal bunch, according to the ability weights of formula (1) and (2) calculating member node, the node of selective power maximum weight, as new bunch of head, goes to step 6.2;

Step 6.2, former bunch of head broadcasts new bunch of header and gap information, and in bunch, member upgrades corresponding information, then enters the transfer of data stabilization sub stage, goes to step 6.3;

Step 6.3, judges whether the local bunch head renewal rewards theory completing k wheel, is go to step 7, otherwise goes to step 6.1.